Why doesn't Intel make dedicated graphics cards?
Why doesn't Nvidia make desktop CPUs?
How good would they be if they did?
>>51495664
>Why doesn't Intel make dedicated graphics cards?
They don't have the necessary IP to compete with AMD and Nvidia. Trying to break into that field would be extremely costly, and wouldn't guarantee them a return on their investment.
Intel is a well run business, and they don't take any uncalculated risks.
>Why doesn't Nvidia make desktop CPUs?
They don't have a standing X86 license, and again they don't have the necessary IP to compete with AMD and intel.
HAHAHAHAHAHAHAHA nVidia can't even compete in the GPU market that is their home territory, they would be completely lost trying to put together a competitive CPU...
>>51495664
>Why doesn't Intel make dedicated graphics cards?
They tried once, it failed miserably. Their GPU (Larrabee) got its ass kicked by G80 (Geforce 8800GTX) despite being 2 years newer than the Nvidia one they were looking to compete with.
But they do.
https://en.wikipedia.org/wiki/Intel740
>>51495664
>http://www.intel.com/content/www/us/en/processors/xeon/xeon-phi-detail.html
>http://www.nvidia.com/object/tegra-4-processor.html
>>51495831
>arm
>real cpu
kek
>>51496098
How is it not?
>>51495682
>IP
The term “intellectual property” carries a hidden assumption—that the way to think about all these disparate issues is based on an analogy with physical objects, and our conception of them as physical property.
When it comes to copying, this analogy disregards the crucial difference between material objects and information: information can be copied and shared almost effortlessly, while material objects can't be.
To avoid spreading unnecessary bias and confusion, it is best to adopt a firm policy not to speak or even think in terms of “intellectual property”.
Why doesn't mcdonalds start selling mexican food?
>>51498071
Hey, RMS: Fuck off and don't come back.
>>51498098
What he's saying is one hundred percent correct.
>>51498094
The mexicans making it fuck up the McDonald's food enough already let alone throwing in more garbage.
>>51498071
Intellectual property is a reality. That is to say, if Nvidia makes use of Intel's IP without permission, the state will fuck them hard. I find it's best to adopt a firm policy of speaking of the world in terms of how it actually is, not how I wish it would be.
meh. Intel tried a discrete GPU project at one point. They decided to focus on integrated GPUs instead given it's more relevant to their core business. NVIDIA has a CPU - it's the Tegra and it's used in mobile computing. They could put it on a desktop form-factor motherboard if they wanted to, and it might be competitive with some low-end AMD offerings, but who wants that?
Closest thing Intel has to a dedicated GPU is the Intel Phi. It's PCIE card with x86 cores on it.
High end has hundreds of cores and low end is about 50. Each core can do 4 logical threads.
>>51498847
The high end of Knights Landing is only 72 cores.
Try again
Yes, im an idiot, but if gpus are so much faster and with more cores and intels best is a 20 core 40 thread cpu - what makes the difference that a super multi core processor couldnt be a cpu? Is it just the way the information is handled in the processing stage, or is there more to a core than just a "core". I just know its bad to let the magic smoke out of them...
>>51498687
I do, I want to see ARM with a good GPU go mainstream.
>>51499098
GPU cores and CPU cores are totally different. CPU cores are capable of a wide range of different operations, where GPU cores only do a very specific type of math.
>>51499228
Got it. Thanks!
>>51499098
Boils down to instruction sets, unless I'm mistaken
>>51499098
>>51499279
To be more specific, Intel cores are very capable of processing highly branched code because of magick such as branch prediction and switching, and general single-threaded performance. Think "Read this long list of instructions that have dozens of if-else conditions"
GPU cores are good at processing thousands of parallel, but simple calculations. Think "apply this same formula to these 100,000 numbers", usually matrix calculations. If you demand a GPU to perform if-else conditionals too much, GPU performance takes a massive hit.
I write CUDA code for a living and my mantra is always "abstract conditionals to the host, perform calculations on the device".
>>51499361
Awesome! Perfect analogy. Thanks for taking the mystery out of it!